Predicted norovirus resurgence in 2021–2022 due to the relaxation of nonpharmaceutical interventions associated with COVID-19 restrictions in England: a mathematical modeling study

In late January 2020, COVID-19 was first detected in the United Kingdom (UK) [1], and in response to increasing incidence and hospitalisations, the UK entered the first of three national lockdowns on the 23rd of March 2020. The lockdowns and other social restrictions resulted in a reduction in contact patterns in the community, as shown through self-reported contact rates collected using online surveys [2] and other metrics [3]. Subsequently, reports to national surveillance for many endemic diseases, including norovirus, declined across England throughout the rest of 2020 [4]. This reduction in reports of infectious disease has been consistent across many pathogens and observed in other countries where non-pharmaceutical interventions (NPIs) for the control of COVID-19 have been implemented (e.g. Australia [5, 6] and the USA [7, 8]). With the introduction of vaccination against COVID-19 in England, a considerable reduction in COVID-19 hospitalisations and deaths has been observed and this intervention has superseded the blunt tool of NPIs. The UK roadmap out of lockdown details the gradual relaxation of NPIs in 2021 [9], which will likely result in contact patterns more consistent with those seen prior to 2020.

Norovirus is an endemic viral infection present globally across high- and low-income countries with—as yet—no commercially available vaccine [10]. Norovirus infection can be asymptomatic or can result in gastrointestinal disease symptoms such as stomach cramps, diarrhoea and vomiting. In more vulnerable groups, such as the elderly or immunocompromised, norovirus can result in more severe and/or prolonged disease, and affected individuals may require hospitalisation for treatment. A recent analysis of hospital data in England ascribes an average of 40,800 (with an interquartile range (IQR) of 40,500–41,400) hospital episodes where norovirus was the primary diagnosis and a further 61,500 (IQR 58,700–62,500) with norovirus as a secondary diagnosis, meaning that norovirus is a significant public health and economic burden in England [11] and during hospital outbreaks can result in ward closures. Norovirus-associated mortality within Europe has been estimated as 0.2 (95% uncertainty interval 0.1–0.2) deaths per 100,000 persons [12], which is at least 100-fold less than equivalent estimates for all-cause influenza illness [13]. Norovirus is the most common cause of gastrointestinal infections globally and has the highest burden of disease for intestinal infections in the UK [14]. Infections and outbreaks occur more frequently in the winter months, and in England are monitored through multiple national surveillance systems (e.g. [15]) to ensure that unusual activity is detected and that alerts to local health authorities are made for preventive actions. The probability of symptomatic disease is also influenced by host genetic factors [16], the infectious dose of norovirus [17] and previous exposure to infection. Norovirus genetic diversity is characterised by classifying isolates into genogroups, genotypes and strains (also termed variants) [18]; since 2012, the GII.4 genotype 2012 strain (shortened to ‘GII.4/2012’) has dominated in reports of symptomatic disease in England and globally. Prior to the GII.4/2012 strain, population strain replacement occurred every few years and was consistent across wide geographies. During these strain replacement events, increased norovirus activity in the first year is typically observed, resulting in increased strain on health services [19]. While each dominant strain is termed dominant because it is isolated from a majority of cases in most settings, there is a considerable genetic diversity of the remaining isolates, and the evolutional impact of this remains unclear. The majority of norovirus infection is considered to occur via community transmission, and consequent changes in community contact patterns will likely influence norovirus transmission. However, norovirus has multiple routes of transmission; for example foodborne transmission is thought to contribute to approximately 16–35% in the UK, based on microbial risk assessments and detailed case investigations [20].

Here we illustrate through modelling the impact of COVID-19 NPIs used throughout 2020–2021 on the dynamics and incidence of norovirus disease in the community, leading on from a recent report of a reduction in norovirus activity [4]. We provide indicators of surveillance to inform what early warning of a norovirus resurgence may look like.

The age-specific incidence of norovirus observed in England is best replicated assuming that primary infection is symptomatic, the duration of asymptomatic infectiousness is 15 to 20 days, and that asymptomatic infection is considerably less infectious than symptomatic infection (Fig. 2A and Table 1). Simulations that correspond to these observations resulted in a shedding prevalence of less than 1%, the proportion of the population recently recovered from an infection ranging from 24.54 to 30.55% and an R₀ between 1.81 and 2.01. The age-specific incidence of symptomatic infection is highest in children aged <5 years, and incidence declines with age. When we compare this modelled incidence to the weekly SGSS 5-year average (from norovirus year 2014/15 to 2018/19), a further 27% reduction in reporting is needed for cases to be equivalent (Fig. 2B). Other models with different assumptions had a poorer fit to the data and were not taken further in the analysis.

From March 23, 2020, (‘lockdown 1’) until September 4, 2020, (the ‘School reopening’ stage), the rate of infection for norovirus is sufficiently low that new infections are rare (Fig. 3A), and an increase in population susceptibility is observed (Fig. 3B). This period corresponds with minimal cases reported to the laboratory surveillance SGSS system (Figure S1). This period corresponded with an R₀ for norovirus ranging from 0.43 to 0.77, meaning that transmission could not be sustained. The period between September 4, 2020, to January 5, 2021, corresponds to a higher average number of contacts within the population (up to 7.79 in the average number of (unweighted) contacts), an increase of incidence in the model, and outside of the second lockdown and the Christmas period, R₀ was above 1.00 (see ‘Additional file 4’). This increase in incidence was not observed in the SGSS system. The subsequent ‘lockdown 3’ period corresponds to a reduction in contacts (from 6.61 to 3.47), and the rate of infection falls to low levels again, until schools are re-opened on March 9, 2021 (Fig. 3A). By this time point, the model estimates that the proportion of the population susceptible to symptomatic infection has risen from 54 to 59% (a 9% increase) as a result of waning immunity (Fig. 3B, age-specific values given in Additional file 1: Table S1). Subsequently, model scenarios predict a rise in the rate of infection and a resurgence of norovirus in the community resulting in an annual incidence of cases up to 2-fold higher than simulations prior to 2020 (Fig. 4A, full details in Additional file 1: Table S1). This time period also corresponds with an R₀ above 1.00, where the contact data collected while schools correspond to an estimated R₀ of 1.28. However, this prediction is dependent upon assumptions on mixing patterns in the general population; if adult mixing is assumed to be 80% of pre-COVID levels, an increase in incidence is not predicted. Instead, the rate of infection increases but at a lower level, resulting in a gradual increase in susceptibility in the population.

We also tested the impact of model assumptions regarding under-reporting and the duration of shedding (Figure S2). Assumptions about under-reporting affect the rate of infection; if incidence was under-reported in Harris [31], this will result in a higher rate of infection in the community and the potential for a more rapid change in incidence. Assumptions on the duration of shedding have a lesser impact; simulation assuming a shorter duration of shedding suggests a higher incidence in the 2021/22 norovirus year (Additional file 1: Table S1).

The reduced reporting of norovirus cases and outbreaks that were observed in 2020 as a consequence of COVID-19-associated NPIs has been described within a number of countries, including England [4], Australia [5] and the USA [7, 8]. Our modelling study quantifies the potential unintended consequences of the build-up of individuals susceptible to norovirus infection in England, with epidemic potential for the autumn/winter of 2021 and 2022. Norovirus is an example of an endemic infectious disease where changes in contact rates can affect both the short and long-term incidence of symptomatic disease. This resurgence is likely to further increase ‘winter pressures’ within the NHS that may lead to more hospital admissions due to norovirus, hospital outbreaks of norovirus with bed-days lost as beds are kept empty for infection prevention and control, and treatment delays for non-norovirus patients [11, 35]; therefore, preparations for mitigating the effect of an increase in incidence are essential. This is especially true at the current time where there has been considerable disruption to healthcare services due to the COVID-19 pandemic and there are a record number of 5 million people on waiting lists to receive hospital care following the impact of COVID-19 in England [36]. While there are several caveats of the modelling to consider in the interpretation of this analysis, this is the first analysis of disease dynamics for norovirus affected by changes in NPIs.

More broadly, the unintended consequences of NPIs are likely to be widespread across other endemic diseases (e.g. respiratory syncytial virus and influenza 37) where incidence pre-COVID-19 was limited by populations being largely immune to infection [38, 39]. A challenge in the interpretation of the likely future trajectory of incidence is the quantification of how NPIs impact physical contacts in the community and how this impacts the transmission of infectious disease. Here we make use of the longitudinal cross-sectional study of contacts, which has been instrumental in predicting the incidence of COVID-19 [40], and show that these data can be very useful for quantifying dynamics of endemic diseases in addition to COVID-19. Continued collection of these data and incorporation into models of infectious disease will be essential in informing our understanding of disease transmission as the pandemic continues to unfold. It is likely that the impact of NPIs on norovirus, respiratory syncytial virus and influenza was more pronounced than for SARS-CoV-2 because the effective reproduction number (R_e, equivalent to R₀ multiplied by the proportion susceptible) for these endemic pathogens is close to 1.00. Theoretically, the NPIs are designed to limit contact rates, which will reduce R₀ and in turn R_e. At least in the period of time that has been experienced, this will limit transmission and disease incidence, as we have observed. The impact of further restrictions should still be effective, but with increases in susceptibility, the indirect effect of NPIs on endemic diseases could become less effective with time. Our analyses make the implicit assumption that contacts described within surveys correlate with contacts relevant for infectious disease transmission where the primary mode is direct contact, while this is apparent for infections such as influenza and SARS-CoV-2, there is less evidence for norovirus. These assumptions explain why the rate of infection is predicted to increase when contacts correspond to an R₀ above 1.00. The reported incidence of norovirus in the coming months, the changing behaviours of the community in contacts and infection prevention, and the actions public health officials will take to prevent a large epidemic, will be the test of this hypothesis. As the COVID-19 pandemic evolves, potentially including more transmissible and vaccine-evading variants, NPIs that include the ongoing working-from-home recommendation through to school closures and lockdowns, remain possible. A major assumption of our model, which reflects the current understanding of norovirus epidemiology, is that contacts with and between children under 5 years are the main driver of norovirus transmission. Consequently, any changes in contact with this age group will have a large effect on norovirus dynamics.

In the analysis, the incidence of norovirus reported was replicated well in the simulations, with the exception of the expected increase in cases at the end of 2020. This potentially suggests that the rate of infection assumed in the model is higher than is experienced in the community, despite the fact that the rate of infection was fitted to community data. Alternatively, the effect of reduced to the SGSS due to changes in healthcare-seeking behaviours and reduced laboratory testing of samples obtained may have impacted on the comparator data. This observation means that some caution should be applied to the findings presented here as the model predictions are sensitive to assumptions about contact rates and the probability of transmission, which remain uncertain. Further work will explore this in more detail through the use of additional contact surveys and exploration of other datasets that can improve our understanding of norovirus transmission in the community.

The majority of norovirus disease occurs within the community; there is a significant disease burden resulting in school closures and days of work lost (especially as infection can be experienced relatively frequently) [41]. Hygiene measures, such as handwashing with soap, enhanced environmental cleaning and staying at home when ill, have been part of the usual recommended practice to reduce norovirus transmission, but the COVID-19 pandemic has indirectly enforced the adherence to these measures. It is possible that beyond 2021, these measures may remain as usual practice in the short to medium term, which may well mean that the incidence of infectious diseases such as norovirus is lowered compared with historical norms. An additional challenge of norovirus largely being a community illness is that norovirus surveillance is affected by both under-ascertainment and under-reporting, resulting in considerable uncertainty in true estimates of burden. With an increase in community transmission of norovirus, we would expect an increase in incidence, including severe disease in institutional settings (such as care homes for the elderly). However, much of the COVID-19 interventions targeted at institutional settings may also limit the incidence of norovirus; the impact of these interventions on norovirus warrants investigation. While we compared the model output to cases reported into the SGSS, surveillance is biased towards cases within hospitals, which will be a combination of community and hospital-acquired infections. For this reason, we have fitted to data from a community cohort study and used the SGSS data as a validation of the model estimates. The data from IID2 is over 10 years old, and may not fully represent incidence experienced in recent years, due to strain replacement in 2012 and hypothesised higher incidence of infection. Indeed, a recent community cohort study in the Netherlands report an incidence of symptomatic norovirus of 339.4 cases per 1000 person-years [42], which is higher than that reported in the IID2. We aimed to overcome the weaknesses in data by having multiple models with varying assumptions to cover a wide range of plausible present-day scenarios of norovirus dynamics. We identified a difference between model predictions and SGSS of 73%, which is well within a plausible range.

A further complexity in norovirus epidemiology is the possibility of strain replacement in the coming months or years; the GII.4/2012 strain has been in circulation in England since 2013 and could continue to dominate, but the emergence of other variants also has the potential [43]. The likelihood and how a new variant may impact norovirus dynamics are currently uncertain and are likely to depend on the relative virulence of a new strain and the heterotypic immunity from infection with the GII.4/2012 strain. Further work will include consideration of strain-specific immunity profiles within the community to understand the impact of multi-strain dynamics and the potential impact of vaccination.

This modelling study suggests that norovirus incidence in the community is likely to remain approximately similar to or substantially increase beyond what has been experienced in years prior to 2020. The results are sensitive to future contact patterns in the community, and the adherence to preventive actions, which will affect the probability of transmission. The lower incidence of norovirus reported in surveillance aligns with model assumptions on reduced contact rates from March 2020 to June 2021 and is consistent with an increase in population susceptibility. The short- to long-term impact of this increased susceptibility places populations at risk of norovirus disease, but the scale of the impact remains to be seen. Continued investment to maintain robust national surveillance systems will remain critical to enable measures to limit the impact of these resurgences and provide essential information to public health bodies to support the implementation of preventive actions.